# Open-vocabulary Segmentation on PACO

Source: [Open-vocabulary Segmentation on PACO](<https://www.sota2.com/research/sota/open-vocabulary-segmentation-on-paco>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-18T08:48:20.436Z

Evaluation of Open-vocabulary Segmentation performance on the PACO dataset, reporting metrics such as AP\_unseen, AP\_seen, Rd, and AND-Eff.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Open-vocabulary Segmentation |
| Dataset (including any stated split) | PACO |
| Metric | AP Unseen |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [AP Unseen](<https://www.sota2.com/research/sota/open-vocabulary-segmentation-on-paco.md?metric=AP%20Unseen>): Higher is better; unit: PERCENT.

- [AP Seen](<https://www.sota2.com/research/sota/open-vocabulary-segmentation-on-paco.md?metric=AP%20Seen>): Higher is better; unit: PERCENT.

- [Rd](<https://www.sota2.com/research/sota/open-vocabulary-segmentation-on-paco.md?metric=Rd>): Higher is better; unit: SCALAR.

- [AND-Eff](<https://www.sota2.com/research/sota/open-vocabulary-segmentation-on-paco.md?metric=AND-Eff>): Higher is better; unit: SCALAR.

## Recorded results

| Method | AP Unseen | Result date | Report paper |
| --- | --- | --- | --- |
| Ours | 39.2 | Not recorded | 6a0a9d2bdbc836b87d82b4c4 |
| SAM3 | 37.9 | Not recorded | 6a0a9d2bdbc836b87d82b4c4 |
| X-Decoder | 33.6 | Not recorded | 6a0a9d2bdbc836b87d82b4c4 |
| CAT-Seg | 32.8 | Not recorded | 6a0a9d2bdbc836b87d82b4c4 |
| FC-CLIP | 29.7 | Not recorded | 6a0a9d2bdbc836b87d82b4c4 |
| OVSeg | 26.3 | Not recorded | 6a0a9d2bdbc836b87d82b4c4 |

### Ours

AP Unseen: 39.2. Result ID: 6a0ad19c47f34ce924f10bbc.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2bdbc836b87d82b4c4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a0a9d2bdbc836b87d82b4c4.md>)

Recorded evaluation configuration: {"Backbone":"SAM3"}. This record does not assert independent reproduction.

### SAM3

AP Unseen: 37.9. Result ID: 6a0ad19c47f34ce924f10bbb.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2bdbc836b87d82b4c4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6923e69bd6b0dd334292f0ad.md>)

[Implementation](<https://github.com/facebookresearch/sam3>)

Recorded evaluation configuration: {"Backbone":"SAM3"}. This record does not assert independent reproduction.

### X-Decoder

AP Unseen: 33.6. Result ID: 6a0ad19c47f34ce924f10bba.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2bdbc836b87d82b4c4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6e3d1e209131cea3e910.md>)

[Implementation](<https://x-decoder-vl.github.io/>)

Recorded evaluation configuration: {"Backbone":"Focal-L"}. This record does not assert independent reproduction.

### CAT-Seg

AP Unseen: 32.8. Result ID: 6a0ad19c47f34ce924f10bb9.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2bdbc836b87d82b4c4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa72b11e209131cea57ef4.md>)

[Implementation](<https://ku-cvlab.github.io/CAT-Seg/>)

Recorded evaluation configuration: {"Backbone":"CLIP ViT-B"}. This record does not assert independent reproduction.

### FC-CLIP

AP Unseen: 29.7. Result ID: 6a0ad19c47f34ce924f10bb8.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2bdbc836b87d82b4c4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa71291e209131cea4f931.md>)

[Implementation](<https://github.com/bytedance/fc-clip>)

Recorded evaluation configuration: {"Backbone":"ConvNeXt-L"}. This record does not assert independent reproduction.

### OVSeg

AP Unseen: 26.3. Result ID: 6a0ad19c47f34ce924f10bb7.

[Reporting paper](<https://www.sota2.com/research/paper/6a0a9d2bdbc836b87d82b4c4.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa6ef21e209131cea4311d.md>)

[Implementation](<https://jeff-liangf.github.io/projects/ovseg>)

Recorded evaluation configuration: {"Backbone":"CLIP ViT-L"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.
